One-Click Live Demo
The fastest way to feel AIRI is the public playground:
Runs the latest main build hosted by the core team.
Log in with Discord to let the bot join your own server instantly.
2️⃣ Prompt-Driven Cloud Deploy (Vercel + Railway)
Copy-paste this block into ChatGPT / Claude and watch it generate a full deployment plan for you.
Prompt
"You are DevOpsGPT. Deploy the open-source project **AIRI** so I can share a link with friends.
REPO: https://github.com/moeru-ai/airi
TARGET CLOUDS:
• Frontend (stage-web) → Vercel (build with pnpm)
• Discord / Telegram services → Railway (Node 20)
• Ollama LLM server → Railway (docker)
STEPS TO OUTPUT:
1. Vercel import command with environment variables DISCORD_CLIENT_ID, API_BASE_URL.
2. Railway CLI script for each service with proper secrets.
3. Health-check URLs.
4. DNS advice so subdomain `chat.mydomain.com` points to Vercel.
Only output runnable bash/JSON snippets, no explanations."The assistant will return ready-to-paste
vercel.json,railway.json, andbashcommands. Execute them in your terminal and you’ll have a shareable URL in ~5 minutes.
3️⃣ Self-Host on Your Box (macOS / Linux)
Bash
# Clone official mirror (includes submodules)
git clone https://github.com/moeru-ai/airi.git && cd airi
# Install everything
pnpm i
# Build native plugins once (Whisper, VAD, window)
pnpm run build:native # wraps `cargo build --release` for crates/
# Launch desktop pet
pnpm dev --filter stage-tamagotchi
# (Optional) Launch bots in new terminals
pnpm dev --filter discord-bot # needs DISCORD_BOT_TOKEN
pnpm dev --filter telegram-bot # needs TELEGRAM_BOT_TOKENDocker Compose (All Services + Ollama)
Bash
# Pre-build Rust for smaller image
pnpm run build:native
# Bring everything up
docker compose -f docker-compose.full.yml up -d- Edit
.envfor LLM model (LLM_MODEL=llama3:8b) and provider keys. - Ollama model layers will download on first run (~4 GB).
Common Pitfalls
| Symptom | Fix |
|---|---|
Module not found: ort | cargo build --release -p tauri-plugin-ipc-audio-transcription-ort then rerun |
| Bot silent in Discord | Enable MESSAGE CONTENT INTENT and VOICE in Discord dev portal |
| GPU not used by ONNX | Set ORT_ENABLE_OPENCL=1 or CUDA_VISIBLE_DEVICES=0 before launch |
Related Technical Files
| Role | Path | Notes |
|---|---|---|
| Desktop app entry | apps/stage-tamagotchi/src-tauri/src/main.rs | Enables transparent window |
| Full compose file | docker-compose.full.yml | Spins up bots + Ollama |
| Railway template | deploy/railway.json | Example service spec |